1st International ICST Workshop on New Computational Methods for Inverse Problems

Research Article

Searchlight CT: A new reconstruction method for collimated X-ray tomography

  • @INPROCEEDINGS{10.4108/icst.valuetools.2011.245816,
        author={Robert  Azencott  and Anando  Sen  and Bernhard G.  Bodmann and King C.  Li and Demetrio  Labate and Xiaobo  Zhou},
        title={Searchlight CT: A new reconstruction method for collimated X-ray tomography},
        proceedings={1st International ICST Workshop on New Computational Methods for Inverse Problems},
        publisher={ACM},
        proceedings_a={NCMIP},
        year={2012},
        month={6},
        keywords={ACM proceedings X-ray Transform Radon Transform Tomography Local Reconstruction},
        doi={10.4108/icst.valuetools.2011.245816}
    }
    
  • Robert Azencott
    Anando Sen
    Bernhard G. Bodmann
    King C. Li
    Demetrio Labate
    Xiaobo Zhou
    Year: 2012
    Searchlight CT: A new reconstruction method for collimated X-ray tomography
    NCMIP
    ICST
    DOI: 10.4108/icst.valuetools.2011.245816
Robert Azencott 1,*, Anando Sen 1, Bernhard G. Bodmann1, King C. Li2, Demetrio Labate3, Xiaobo Zhou2
  • 1: 608 PGH, University of Houston 4800 Calhoun Road Houston, Texas 77204
  • 2: The Methodist Hospital Research Institute 6565 Fannin Street Houston, Texas 77030
  • 3: 694 PGH, University of Houston 4800 Calhoun Road Houston, Texas 77204
*Contact email: razencot@math.uh.edu

Abstract

The method presented in this paper aims to reduce the overall radiation exposure from X-ray CT scans when only the reconstruction of a region of interest is required. To achieve this goal, the Searchlight CT algorithm restricts the acquisition to X-rays passing mainly through the region of interest. The algorithm uses an iteration of the X-ray transform and of a regularized inverse, which converges rapidly and for which generic conditions of converges are provided. The performance of the Searchlight CT algorithm is illustrated on both phantom and experimental data.